[Building Sakai] Tool usage analysis

[David Adams]

On February 14, Martin B. Smith wrote:
> I'd be interested in these scripts too. IIRC, Michigan had some scripts
> that analyzed load balancer log output; I'd love it if I could find some
> to run against Apache 'combined' format logs. I've been planning to
> write some myself, but haven't gotten to it yet.

Martin et al,
I've developed some scripts that combine the httpd access_logs from all our app servers, sort them by request time, and do some post-processing and cross-referencing to figure out what tool, site, and user made each request, and then adds those fields to the resulting combined log, along with simplifying the user agent string.

It's definitely become a critical tool in our arsenal for troubleshooting as well as for reporting. Unfortunately the script relies heavily on some intricacies particular to Virginia Tech's configuration and wouldn't be immediately generalizable. However, if there's interest in a simpler, more general tool, I can probably cut out the VT-specific elements.

A couple things I will say about the general problem of tool-usage reporting: HTTP logs, especially from Sakai, are going to sometimes be a poor proxy for comparing tool usage between tools. One HTTP request in a tool like Samigo might represent an hour's worth of "usage", whereas a five second single-page visit to Gradebook2 could result in dozens of hits. So be sure you understand a tool's usage profile before you compare activity between disparate tools using just raw hit counts.

-dave
-- 
David Adams
Director, Learning Systems Integration and Support
Virginia Tech Learning Technologies